Chinese University Students’ Attitudes Towards China English and Teaching China English: Influential Factors

From the 2000s, attitudinal studies on “China English (CE)” and incorporating CE into English Language Teaching (ELT) have emerged because of CE’s legitimacy as a variety of “New English” and its pedagogical implications being widely recognised. However, no settled conclusions have been made on students’ attitudes towards CE, which could also be constantly shifting. Furthermore, the reliability and validity of existing attitude questionnaires on CE and teaching CE have not been tested, and the impact factors of attitudes toward CE and teaching CE have seldom been explored quantitively. Given these research gaps, the present study investigates students’ attitudes towards CE and the idea of incorporating CE into university ELT. The methods include a questionnaire and statistical approaches such as exploratory factor analysis and factorial MANOVA. The results suggest that the CE and teaching CE scale comprises nine factorial dimensions with acceptable reliability and validity. Furthermore, Chinese university students hold mixed attitudes towards CE and generally negative attitudes towards teaching CE. The results have also revealed that students’ academic discipline and understanding of the “World Englishes” concept are statistically significant predictors of their attitudinal response in some areas. Based on these findings, the study suggests that it is still too early for CE to be taught in the classroom, but the discussions around this topic should not stop. At the same time, various endeavours could be tried to lay the foundation for CE to be incorporated in ELT. Recommendations for future attitudinal studies on CE and teaching CE are also proposed.


Introduction
Over the last three decades, a number of localised or indigenised varieties of English have been observed worldwide owing to the global spread of English, and the term "World Englishes" has been extensively employed (Bolton, 2012). Although scholars have also used a range of other terms such as "Global Englishes," "International Englishes," and "New Englishes," all these research paradigms view English as a pluricentric notion and focus on the global ownership of English independent of native

Current attitudinal studies on CE and teaching CE
As a key concept in applied linguistics and the discussion of CE, attitudes towards CE and teaching CE have been researched by many scholars during the last two decades in the context of the Chinese mainland (e.g., Edwards, 2017;Fang, 2017a;He & Li, 2009;Kirkpatrick & Xu, 2002;Pan, 2019;Wang, 2015;Xu, 2010). However, most of these studies have mainly been restricted to debates over whether it exists as opposed to "Chinglish," which is described as interference from Chinese during the process of English learning rather than a new English variety (Fang, 2017b;Xu, 2010). Therefore, it is still important and necessary to research CE attitudes based on its acceptance as an established variety. Among the existing attitudinal studies, the most relevant one to the current study was conducted by He and Li in 2009. In that study, they echoed the pioneering 2002 attitudinal research conducted by Kirkpatrick and Xu and studied the attitude of university teachers and students towards CE and teaching CE within the university curriculum in the Chinese mainland. After surveying, they found that CE was not widely accepted by teachers and students who still generally favoured so-called "Standard English" (e.g., American English) over CE, similar to the findings of Kirkpatrick and Xu (2002). Despite its substantive contributions, flaws exist in their research methodology. For instance, the cross-validation effect of their "triangulation" research design could not be promised because their matched-guise experiment of the questionnaire and interviews only tested students' attitudes towards the phonic features of CE other than all four levels of CE linguistic features (i.e., phonology, lexis, syntax, and discourse pragmatics). Moreover, the idea of using "triangulation" to guarantee reliability and validity has been under criticism as different research methods are argued to construct different realities (Silverman, 2020). Hence, in order to be more precise about students' attitudes towards CE and teaching CE, more rigorous research methods should be employed. In other attitudinal studies on CE and teaching CE, more focused topics are chosen. For example, Wang (2015) examined the attitudes of university teachers and students towards attitudes of university teachers and students towards specific linguistic features of CE, and the findings suggested that all four levels of CE features received low acceptability among the research group. In another study that focused on students' attitudes towards their own CE accents instead of all CE features (Fang, 2017a), the majority of the participants showed unsatisfactory feelings about their CE accents due to the pervasive native-speakerism in ELT. However, while negative attitudes were dominant in different studies employing different approaches, positive attitudes have also been discovered in some recent research. For instance, Pan (2019) found that the willingness of acceptance was expressed by 26 out of 30 students surveyed and students also showed positive attitudes to the Chinese cultural features of CE (e.g., transliteration words such as "the great revival of the Chinese nation 中 华 民 族 伟 大 复 兴 "). Moreover, students felt proud of CE for its lexical influences on the "central" varieties in Kachru's Three Circle model. For example, "lose face 丢 脸 " , a semantic transfer idiom deriving from Chinese now has a lexical entry in the Merriam-Webster dictionary (Merriam-Webster, n.d.). To sum up, a complex and even internally contradicted picture of how students react to CE and the idea of incorporating it into university ELT can be observed from the existing research findings. In addition to the current methodological imperfections and unsettled conclusions, scholars have also emphasised that it is worthwhile to re-conduct attitudinal research with a comparable cohort every few years because attitudes are constantly shifting (Kirkpatrick & Xu, 2002;Wang, 2015). Therefore, given these research gaps, the current project is dedicated to discovering university students' up-to-date attitudes through more comprehensive research instruments and rigorous statistical measures.
In addition, in parallel fields of "language attitude" studies, not only attitudes but also their impact factors have been extensively studied. For instance, Wei et al. (2020) chose gender, grade, language proficiency, and perceived language learning difficulty as their focal variables and found gender was a statistically significant predictor of Chinese Mongolian students' attitudes towards trilingualism. However, few studies have investigated the influencing factors of CE attitudes and hence, the current study also endeavours to study these influences. Among the two traced studies, He (2020) found that Chinese students of different disciplines and genders showed significantly different attitudes towards CE and "Standard English." Specifically, male students were more willing to be identified clearly as Chinese while speaking English. Therefore, gender and major are selected as the first two focal variables of the current project. Furthermore, Wang and Jenkins (2016) discovered that Chinese students' experience of using English as a Lingua Franca could resolve their negative attitudes towards language varieties other than "Native English" caused by the concerns of their intelligibility. Hence, university type (carrying out English Medium Instruction or not) and understanding of the "World Englishes" concept are also chosen as predictor variables to be studied in this project.
To conclude, to fill the current research gaps and further study students' attitudes, three research questions were asked in the current research: RQ1: What are university students' attitudes towards CE? RQ2: What are university students' attitudes towards incorporating CE into university ELT? RQ3:

Instrument
In the current project, a quantitative method was adopted. The questionnaire used in the current report consisted of two sections.
To what extent do the selected variables (i.e., gender, academic discipline, university type (carrying out English Medium Instruction or not) and understanding of the "World Englishes" concept) affect university students' attitudes towards CE and incorporating CE into university ELT?
The first part of the questionnaire surveyed the selected variables (see Appendix A): students' gender, university type, academic discipline, and understanding of the "World Englishes" concept. It is important to note that the questionnaire asked students to submit their university names in a text form, and then their university types were manually categorised by the researcher.
In the second part of the questionnaire, 31 items were used to test students' attitudes towards CE and incorporating CE into university ELT. In the form of a Likert scale, students were asked to express their degree of approval to the 31 items with a score. A higher number indicated a higher level of approval, with the lowest number, 1, standing for "totally disagree," and the highest number, 6, standing for "totally agree." Most of the items were adapted from He and Li (2009), but changes had been made to improve their reliability and suit the purpose of the current research. For instance, "American and British English" was used in the questionnaire instead of "Native Speakers" to avoid any definitional ambiguity. Additionally, questions containing technical linguistic terms such as "The variety of English in China should have its own linguistic features at the levels of phonology, lexis, syntax and discoursepragmatics" were simplified. To better answer the research questions of the current project, some items were omitted, and others were added. Specifically, as the current study built on the ground that CE is an established variety of English, the definition of CE was given to the participants and questions surveying students' attitudes towards the name of CE were deleted. In contrast, more statements about ELT and possible reasons behind students' attitudes were included. Parts of the new questions drew inspiration from Xu's (2010) questionnaire, which also surveys CE attitudes, and findings from Pan (2019), which provides a reasonable explanation of students' attitudes. The form of a 5-item Likert scale in He and Li's research was changed into 6-item to guide respondents to answer at either end of the scale (Rasinger, 2008), and negative constructions (i.e., "no" or "not") were avoided in the questionnaire to minimise the chances of misreading (Dörnyei, 2003).
In addition, the questionnaire was piloted among 10 university students. Based on their feedback, three more major revisions had been made: all items were presented only in Chinese in the distributed version of the questionnaire; the wording and phrasing of a few items had been changed; more examples of CE were provided to help the respondents understand and answer the questions.

Participants
After an approval was received from the ethics committee of the author's affiliation, the questionnaire was distributed through the Chinese online survey platform, Wenjuanxing. In total, 155 responses from Chinese students were collected. Among all the valid respondents, 74.8% were female (n = 116); 23.2% were male (n = 36). This gender distribution was expected since the dominance of female participants is typical in Web-based foreign-language-related surveys (Wilson & Dewaele, 2010). At the time of the study, 69.7% of the participants were students studying in Xi'an Jiaotong-Liverpool University (n = 108), an international joint university located in east China, while the other participants were from different universities across the Chinese mainland, such as the South China University of Technology. Besides, English major students were included in the current study.

Research questions and analytical strategies
The questionnaire data was imported into the Statistical Package for Social Sciences (SPSS) 26.0 to perform statistical analyses. Before answering RQ1 and RQ2, the internal structure of the CE and teaching CE scale was studied using exploratory factor analysis. When performing the analysis, principal components analysis was selected as the factor extraction method, and the direct oblique rotation with Kaiser normalisation assuming correlations among factors was used as the rotation method since factors in the field of social science tend to intercorrelate with each other (Field, 2009). Because the direct oblique rotation method was used, both pattern matrix (i.e., the regression coefficients for each variable on each factor) and structure matrix (i.e., the correlation coefficients between each variable and factor) were inspected according to Field's suggestion (2009). As the variables derived from factor analysis only statistically make sense, the questionnaire items constituting each factor were decided based on their factor loading scores in two matrixes and practical meanings. The cut-off point of .40 for factor loading scores was adopted (Field, 2009). After extracting the factors, the construct validity of the CE and teaching CE scale was informed from the percentage of variance explained (Field, 2009), and its reliability was measured with the most frequent index, Cronbach alpha (Derrick, 2016). Reliability analyses of nine factors were separately conducted as Cronbach suggested (as cited in Field, 2009). Furthermore, the factor scores were generated accordingly using the regression method instead of the weighted average method. Although both methods can produce factor scores that are strongly correlated with the original data to maximize validity and obtain objective estimates of the true factor scores, the regression method is more pinpoint because it uses factor score coefficients (B) instead of raw factor loadings (b) as weights like the weighted average method does (see Equation 1) (Field, 2009). The factor scores were used as data in the following multivariate analysis of variance analysis (MANOVA).

Equation 1
The Formula of the Regression Method Factor i =B 1 Variable 1i + B 2 Variable 2i + ⋯ + B n Variable ni + ε i B=R -1 A Note: R -1 : the inverse of the original correlation or R-matrix; A: original data RQ1 and RQ2 were answered through descriptive statistics such as frequency, means and standard deviations (SDs). The results of items adapted from He and Li (2009) were also visually inspected and compared with their results to determine if students' attitudes had changed. A set of one-sample t-test was run using the scorings from He and Li as test values to find out whether the differences discovered were statistically significant. Before conducting the one-sample t-test, data collected by He and Li using a 5-point Likert-scale questionnaire were converted using a coefficient of 6/5 because a 6-point Likert scale was used in current research.
RQ3 was addressed through a factorial MANOVA test. After the MANOVA analysis, follow-up ANOVAs and Post-hoc LSD tests were used to find the exact differences in students' attitudes. The interpretation of MANOVA statistics was based on a combination of Θ (eigenvalue), F ratio (variation between sample means), and p-value (statistical significance level), the partial η 2 (effect size).
In the current research, the effect size were reported as long as it applied to the statistical tests performed as it expresses the magnitude of a relationship and, hence, is more critical than the p-value in quantitative studies (Loewen & Plonsky, 2016). It was interpreted using the benchmark of Cohen (as cited in Leech et al., 2014). Exact p-values were also reported in the current study except for cases where very extreme p-values reported as p < .05 or p > .05, and all the results were retained to the maximum of three decimal places.

The factorial structure of the attitudes towards CE and teaching CE scale
Before conducting the exploratory factor analysis, its assumptions were first checked. Firstly, one item was deleted because it correlated poorly with all other items (Field, 2009). All other items of the questionnaire had adequate effect size (p < .001, r ≥ 0.3) and were therefore retained. The distribution of data was normal, the homogeneity of variance was checked, and the results of Bartlett's Test of Sphericity indicated that correlations between items were sufficiently large for principal component analysis (χ 2 (155) = 2040.487, df = 435, p < .001). The KMO values calculated from Kaiser-Meyer-Olkin Measure of Sampling was .714, "good" according to Hutcheson and Sofroniou (as cited in Field, 2009, p. 659). Therefore, the dataset was appropriate for factor analysis.
The initial analysis indicated there were nine components with eigenvalues bigger than Kaiser's criterion of 1, while the scree plot showed inflexions in components 1, 2, 3, 4, 5, 6, 7, 9 (see Appendix B). The average of communalities after extraction was 0.6789 while there were 30 variables analysed, close to Kaiser's standard of accuracy guarantee (Field, 2009). Therefore, given the small number of items, the nine-factor solution was accepted based on the convergence of Kaiser's criterion and the visual inspection of the scree plot. Then, based on the factor loadings in both the pattern matrix and the structure matrix, and their practical meanings, the questionnaire items were distributed to the most appropriate factor among the nine, which were named later. The final results were summarised in Table  1 with the percentage of variance explained and Cronbach's α values presented. The nine-factor solution explained 67.887% of the variance found in the dataset, proving that the scale had overall great construct validity. Moreover, it could be found that five subscales had high reliability with Cronbach's α bigger than .7, whereas the other four subscales had relatively low reliability with Cronbach's α smaller than .7. However, Field illustrated that apart from the generally used cut-off point of .7, a lower Cronbach's α could be expected for psychological constructs because of its diverse nature (2009). In other words, some of the CE subscales performed excellently in the reliability test, and the reliability of all CE subscales was sufficiently robust.

Students' attitudes towards CE
Students' attitudinal responses to CE are presented in Table 2. It could be informed by the results of questionnaire item 15, "When I speak English, I want to sound like a native speaker from America or Britain" (mean = 4.61, SD = 1.351), that Chinese university students' desire to sound like "native speakers" was still strong. Among all participants, 33.5% chose the second-highest score of 5 ("agree"), and 30.3% chose the highest score of 6 ("totally agree") for this statement. This result was consistent with the findings of two recent studies that specifically focused on researching Chinese students' accents (Huang & Hashim, 2020;Kung & Wang, 2019). Both demonstrated that most of the Chinese students still preferred "native speaker" accents. However, the "native speaker" concept was problematic and hard to be defined with a unified standard. Many speakers of the "New Englishes" were also "Native English" speakers, although people usually associated "native speaker" with speakers from the U.S. or U.K. (Galloway & Rose, 2015). In the research conducted by Tévar (2020), Chinese students rated CE accent as the third place when it was situated in a group of English varieties, before accents of outer circle Englishes (e.g., Russian, Hispanic, Indian, Nigerian). In that same research, the British accent stood out to be the most favoured accent amongst all accents provided. Similar research findings were discovered by Pan (2019) as well; U.S. accent and U.K. accent were rated as the first and second most preferred accent by Chinese students with a huge lead before CE accent and other types of English accents, including Canadian and French English accent. Therefore, instead of saying students adopt the language ideology of "native-speakerism," it was more accurate to say that most students only recognised English varieties from inner-circle countries of Kachru's three-circle model as legitimate forms of English. This phenomenon might be caused by the intelligibility of different varieties of English to Chinese students. When surveyed about the easiness of understanding in the current research, students indicated that it was easier to understand American or British English than English varieties from Singapore (an outer-circle country) or Japan (an expanding-circle country) both phonologically (item 6, mean = 4.95, SD = 1.021) and textually (item 7, mean = 4.56, SD = 1.157). Moreover, this obvious bias towards inner-circle country accents might also be influenced by the British-and American-Englishbased pedagogic models, which were treated as class norms in China. Fang (2016) illustrated that the English textbooks used in China still mainly adopted the Inner Circle variations, leading students to take American or British accents as the ultimate goal of learning English. The results of current research confirmed that British or American English was the dominant varieties learned by Chinese students both in the classrooms (item 17, mean = 4.99, SD = 0.993) and outside of the classroom by themselves (item 16, mean = 5.11, SD = 0.872).
However, students' perceptions of CE seemed to be changing. When the results were compared with those discovered by He and Li (2009) using one-sample t-tests, students in the current study were less interested in sounding like speakers from America or Britain and more willing to demonstrate their own Chinese identity when speaking English. To be exact, in questionnaire item 15, students' mean score discovered by He and Li was 4.33 on a 5-point Likert scale (converted score = 5.196), meaning students gravitated heavily towards the opinion that "When I speak English, I want to sound like a native speaker from America or Britain." Compared with the mean value of 4.61 discovered in the current 6-point Likert scale, students' attitudes towards this item had become comparatively more positive. The result of the one-sample t-test for it was p < .000, r = .401, indicating a statistically significant variance and small effect size based on Cohen's benchmark (as cited in Leech et al., 2014). In other words, the surveyed population in the current research did show more acceptance of CE accents.

Students' attitudes towards incorporating CE into university ELT
When learning preference was surveyed (see Table 2), students were primarily fond of learning British or American English by broadly agreeing with questionnaire item 18, "When we learn English, we should learn British or American English" (mean = 4.47, SD = 1.083). They disapproved the idea of incorporating well-defined features of CE into the existing teaching model (item 2, mean = 3.24, SD = 1.315) nor the more extreme notion of "China English can replace the existing teaching model" (item 29, mean = 2.47, SD = 1.118).
Furthermore, different from an upsurge in recognising and accepting CE, students' attitudes towards incorporating CE into university ELT were contrastively becoming more conservative compared with He and Li's research results. Only 4.5% of students were totally in favour of bringing in well-defined features of CE into university ELT (item 2, mean = 3.24, SD = 1.315), while 25.5% of the students totally agreed with this idea in the 2009 research of He and Li (mean = 3.66, converted mean = 4.392). The result of the one-sample t-test proved the statistical significance of the difference revealed from comparing means: p < .000, r = .661, a large effect size based on Cohen's benchmark (as cited in Leech et al., 2014). Meanwhile, a higher percentage of respondents in the current research were opposed to "China English can replace the existing teaching model" (item 3, mean = 2.47, SD = 1.118) and "Chinese students should learn China English in addition to American and British English in university English" (item 4, mean = 2.79, SD = 1.294), both tested to be statistically different compared to the results found by He and Li in 2009 with considerable effect sizes (p < .000, r = .591) (p < .000, r = .770). Note: Score 1 refers to "Totally Disagree," 2 refers to "Disagree," 3 refers to "Slightly Disagree," 4 refers to "Slightly Agree," 5 refers to "Agree," 6 refers to "Totally Agree." Apart from showing a noticeable preference for American or British English in ELT, students also firmly believed that their low learning effectiveness was not caused by the adoption of British English or American English as the teaching model (item 1, mean = 2.45, SD = 1.228) and non-native speakers could also learn to speak standard American or British English (item 27, mean = 5.46, SD = 0.75). In other words, students were satisfied with the current curriculum design and believed they could succeed in it. Nevertheless, these two notions were disputed by many scholars in language and education research. For instance, findings from recent research by Fang (2016) showed a mismatch between the objectives mentioned in the ELT course syllabus and the classroom outcome, imposed an unnecessary burden of learning "native" or "authentic" English on students. It is worth mentioning that the surveyed students' overall satisfaction with their own English learning effectiveness was not as positive as students' attitudes towards the current ELT system. The most popular choices were "slightly agree" (43.3%) and "slightly disagree" (27.1%) in questionnaire item 31, "I am satisfied with my English learning effectiveness" (mean = 3.85, SD = .968). This attitudinal response was rather conservative compared with students' firm belief that they can be "Standard English" speakers (item 27, mean = 5.46, SD = 0.75), in a way confirming the scholars' opinions on the unattainability of "Standard English" for most students. In other words, although the current exonormative teaching model might impair students' learning effectiveness, they did not recognise this influence and still supports it entirely.

Students' ambivalent attitudes
The findings from the previous sections demonstrated that the development of students' attitudes towards CE and incorporating CE into university ELT went in two opposite directions. This kind of contradicting phenomenon was also discovered in other parts of the questionnaire responses. Although attitudes toward the linguistic features of CE and its place in university ELT remained negative, students were found to be optimistic about other attitudinal dimensions, for example, the global acceptance of CE. As can be seen from Table 2, the participants generally agreed with the idea that "The variety of English in China is bound to be influenced by the Chinese language" (item 11, mean = 4.92, SD = 1.013). Moreover, questionnaire item 12, "China English is affecting American or British English," was also widely agreed upon by students (mean = 4.15, SD = 1.218). In other words, the participants felt that it was inevitable for CE to be developed in China and that CE was influencing "Standard English," which demonstrated an inverted direction of language transmission in Kachru's three-circle model. However, a significant proportion of them still wanted to sound like a "native speaker." This result coincided with a recent qualitative CE attitudinal study and could be understood with its explanation (Pan, 2019). In Pan's study, students were discovered to believe in two governing forces of language ideology, leading to ambivalence. The first language ideology was positive, such as "CE influences native-speaker English" and "CE demonstrates Chinese culture." In contrast, the second type of language ideology was negative, like "the use of CE is embarrassing" and "I desire to achieve American or British English competence." That is to say, students' positive attitudes towards CE were influenced by the governing ideology of strong cultural identities and national pride, whereas negative attitudes towards CE were affected by the language hegemony of American and British English and students' desire for personal empowerment through learning English.
In the current study, students' national pride was well demonstrated through the strongly positive responses to questionnaire item 13, "I am very proud of being Chinese" (mean = 5.28, SD = 1.029), and 14, "I am happy to let others know that I am from China" (mean = 5.33, SD = .934). Furthermore, the surveyed students also believed in the positive correlation between China's political and economic power and people's attitudes towards CE (item 29, mean = 4.57, SD = 1.162). This kind of positive belief of Chinese students clearly clashed with their negative beliefs about teaching CE. For instance, questionnaire item 3, "China English can replace the existing teaching model" (mean = 2.47, SD = 1.118), was commonly disagreed. This kind of negotiation between language ideologies also led to the heterogeneity of students' attitudes. Questionnaire item 28, "I want people to know that I am from China when I speak English," had the highest SD value of 1.564 among all questionnaire items. To conclude, Chinese students' positive attitudes towards CE were displayed in the language level from an etic perspective, and their negative attitudes were contributed by their personal aspirations from an emic perspective, explaining why even though students in the current study felt more positive about CE generally, they grew to be less interested in learning it in university ELT personally.

The relationship between the focal variables and students' attitudes
To assess the relationship between the focal variables and students' attitudes, a factorial multivariate analysis of variance (MANOVA) was conducted. Based on the results of principal component analyses in this study, nine factors were extracted, but only factors (1), (2), (4), (5), and (9) with Cronbach's α value greater than .7 were analysed to fulfil MANOVA's statistical assumptions. These five factors were treated as the dependent variables in the MANOVA test, with the four focal variables (all categorical) treated as the independent variables.
Before conducting the analysis, the assumptions of the MANOVA test were checked. Firstly, the collective multivariate normality of the dependent variables was proved through checking the assumption of univariate normality for each dependent variable in turn through the Kolmogorov-Smirnov test. All the Kolmogorov-Smirnov tests conducted showed non-significant results (p > .05) except for factor (4), National Pride (p < .05), which was later excluded so that only factors (1), (2), (4), and (9) fulfilled all the assumptions required for MANOVA and entered the final analysis. Levene's test (p > .05) and Box's test (p > .05) both revealed non-significant results, proving that the assumption of homogeneity of variance was met. The inexistence of multicollinearity among the dependent variables was confirmed with a correlation (p < .05).
These results were only partially consistent with those discovered by He (2020). In that research, gender, grade, teacher or student, discipline, and types of university (key university or second-tier university) were treated as the focal variables. The results indicated that female and male students were different on some items, so did students of different disciplines, while only the academic discipline was found to be a statistically significant influencer of students' attitudes in the present study. This discrepancy might be due to the fact that the MANOVA test conducted by He was done on an item level using the original data collected from the questionnaire, whereas the current project conducted the MANOVA test using data converted through the regression method in the factor analysis, which was more perspicacious (Field, 2009). Moreover, the questionnaire was slightly different in some items.
Follow-up ANOVAs revealed statistically significant variations in factor (9), attitudes towards the global acceptance of CE, from students in different academic disciplines (F(4) = 3.367, p = .012, partial η 2 = .101). Moreover, the effect size value belonged to the range of large or larger than typical. Found in the subsequent Post Hoc LSD test, students from natural science majors had more positive attitudes towards the global acceptance of CE than applied science students (p = .06), presenting a new result that had not been discovered in previous studies.
Concerning the effect of students' understanding of the "World Englishes" concept on their attitudes of the four dimensions of CE and teaching CE scale, the follow-up ANOVAs indicated no statistically significant attitude differences with significant p-values and small partial η 2 values, demanding further studies to check on the exact impacts of students' understanding of the "World Englishes" concept on their attitudes towards CE.
Finally, the results of MANOVA indicated that there was only one set of variables, namely the combination of gender, university type (carrying out English Medium Instruction or not) and understanding of the "World Englishes" concept demonstrated statistically significant influence on students' attitudes (Θ = .069, F(4, 118) = 2.021, p = 0.022, partial η 2 = .063), meaning that the linear composite of four dimensions of CE and teaching CE scale differed for students with different combinations of characteristics on the three predictor variables. The follow-up ANOVAs further illustrated that the interaction had statistically significant power on factor (5), attitudes towards Standard English, and factor (9), attitudes towards the global acceptance of China English. However, due to the limitations of MANOVA's statistical power, the exact relationship remained unclear.

Conclusion
The present study has built upon the attitudinal research of He and Li (2009) by partially replicating it in a similar group but a different year with more rigorous statistical approaches and some new research focuses. The CE and teaching CE scale adapted from He and Li is discovered to comprise nine factorial dimensions with great overall validity and acceptable reliability for all individual factors. The questionnaire survey results indicate that university students have mixed attitudes towards CE and generally negative attitudes towards incorporating CE into teaching. Specifically, the current study participants feel proud of CE because of their strong cultural identities but prefer American or British English when learning because of the stigmas attached to CE and their desire for personal empowerment through learning English. Based on the results of MANOVA, students in different academic disciplines possess attitudes that are significantly different in statistics, so do students at different levels of understanding of the "World Englishes" concept. Meanwhile, gender and university type are statistically non-significant in predicting students' attitudinal responses, contradicting the findings of previous research. By providing better instruments and new evidence, future research on attitudes towards CE and teaching CE are solicited.
Based on students' overall conservative attitudes towards CE and resistance to learning CE in university ELT, the time may still be immature for universities in the Chinese mainland to change the currently adopted native speaker model into other teaching methods, such as the exonormative model or the Lingua Franca model. However, given that the study has also shown that students' goal of becoming native-like speakers may be unrealistic and potentially impair their learning, the possibility of different teaching models should continue to be discussed. As students may feel satisfied with the current teaching model merely because it is the only one available and familiar to them, awareness about relevant issues should be raised to education workers in China as a start and students subsequently. Moreover, since the study has also found that understanding of the "World Englishes" concept could impact students' attitudes, "World English" could be introduced in the university ELT curriculum to help students understand the legitimacy of CE if possible. On a higher level, the recognition and acceptance of CE are essential for students' attitudes because of their desire for personal empowerment through learning English. To achieve this goal, criterion (5) of Butler's (1997) theory of determining the existence of a native variety of English, namely, a set of reference works, should be further worked on through the development of dictionaries and teaching materials.
Notwithstanding its substantive and methodological contributions, the present study has two major limitations. Firstly, the data was collected using an online questionnaire, which has both advantages and inherent limitations (Wilson & Dewaele, 2010). Secondly, the results of the current MANOVA analysis are limited by the sample size. Due to the high imperceptibility of the questions that MANOVA answers and the relatively small sample size, parts of the influences of the selected variables may be unexploited in the current research. Therefore, it is suggested for future studies to continue studying the influencers of CE with a larger sample.
Ultimately, the importance of teaching methods is self-evident in China, given the number of people being influenced. The current study does not intend to represent every student but to present the actuality of a group of students to the world and hope to initiate more discussions aiming to bring a future where all users of English can be in one circle instead of three.